Spaces:
Running
Running
Update rag.py
Browse files
rag.py
CHANGED
@@ -1,3 +1,48 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
def get_best_answer(user_input):
|
2 |
user_input_lower = user_input.lower().strip()
|
3 |
|
|
|
1 |
+
import json
|
2 |
+
from sentence_transformers import SentenceTransformer, util
|
3 |
+
from groq import Groq
|
4 |
+
import datetime
|
5 |
+
import requests
|
6 |
+
from io import BytesIO
|
7 |
+
from PIL import Image, ImageDraw, ImageFont
|
8 |
+
import numpy as np
|
9 |
+
from dotenv import load_dotenv
|
10 |
+
import os
|
11 |
+
|
12 |
+
# Load environment variables
|
13 |
+
load_dotenv()
|
14 |
+
|
15 |
+
# Initialize Groq client
|
16 |
+
groq_client = Groq(api_key=os.getenv("GROQ_API_KEY"))
|
17 |
+
|
18 |
+
# Load models and dataset
|
19 |
+
similarity_model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
|
20 |
+
|
21 |
+
# Load dataset (automatically using the path)
|
22 |
+
with open('dataset.json', 'r') as f:
|
23 |
+
dataset = json.load(f)
|
24 |
+
|
25 |
+
# Precompute embeddings
|
26 |
+
dataset_questions = [item.get("input", "").lower().strip() for item in dataset]
|
27 |
+
dataset_answers = [item.get("response", "") for item in dataset]
|
28 |
+
dataset_embeddings = similarity_model.encode(dataset_questions, convert_to_tensor=True)
|
29 |
+
|
30 |
+
def query_groq_llm(prompt, model_name="llama3-70b-8192"):
|
31 |
+
try:
|
32 |
+
chat_completion = groq_client.chat.completions.create(
|
33 |
+
messages=[{
|
34 |
+
"role": "user",
|
35 |
+
"content": prompt
|
36 |
+
}],
|
37 |
+
model=model_name,
|
38 |
+
temperature=0.7,
|
39 |
+
max_tokens=500
|
40 |
+
)
|
41 |
+
return chat_completion.choices[0].message.content.strip()
|
42 |
+
except Exception as e:
|
43 |
+
print(f"Error querying Groq API: {e}")
|
44 |
+
return ""
|
45 |
+
|
46 |
def get_best_answer(user_input):
|
47 |
user_input_lower = user_input.lower().strip()
|
48 |
|